Estimation of Calcium, Magnesium and Sulfur Content in Oil Palm

using Multispectral Imagery based UAV

Muyassar Allam Suyuthi1,*, Kudang Boro Seminar1 and Sudradjat2

1Department of Mechanical and Biosystem Engineering, IPB University, Bogor, Indonesia
2Department of Agronomy Horticulture, IPB University, Bogor, Indonesia

Keywords: Oil Palm, Multispectral Images, Nutrition, UAV.

Abstract: Oil palm is a commodity which contributes to the largest foreign exchange. In Indonesia, the area of oil palm
plantations has a large area compared to other commodities. Proper and efficient fertilization is needed to
reduce production costs. This study aims to estimate the nutrient content of calcium, magnesium, and sulfur
in oil palm plants using UAV-based multispectral cameras. The method used is divided into three stages, data
preparation, pre-processing, and data analysis. At the data preparation stage, the things done are leaf sampling,
sample coordinate points, and multispectral image capture. In the pre-data processing stage the things done
are stitching, georeferencing, and digitizing. The last stage is data analysis like multiple linear regression
analysis to get a model of the relationship of multispectral images and actual nutrition of lab test results. The
results of this study obtained a model for predicting of calcium content Ca = 0.994+0.00723*GREEN-
0.00863*RED EDGE, for magnesium content is Mg = 0.693+0.00531*RED-0.00541*RED EDGE, and for
sulfur content is S = -0.222+0.00338*RED EDGE. The results of overall accuracy using a confusion matrix
of 66.7% in the calcium model, 63.3% in the magnesium model, and 36.6% in the sulfur model.

1 INTRODUCTION plantation area so that the most efficient absorption of

nutrients occurs in plants and can integrate the use of
Oil Palm (Elaeis guineensis Jacq.) is a commodity of mineral fertilizers and oil palm residues and minimize
plantation crops that contributes to the country's environmental impacts associated with excessive
largest foreign exchange because oil palm fertilization such as land degradation (Goh and
commodities have high economic value in Hardter 2003). In addition to maximizing the
agribusiness. Palm oil has benefits and advantages nutrients absorbed, it can also save production costs,
compared to other commodity vegetable oils. because according to Gerendas and Heng (2011) a
Indonesia is the country with the largest palm oil large proportion of the total production costs is
production in the world with a contribution of 44.46% fertilization costs.
of the total world CPO. This contributes to the quite Essential nutrients are essential nutrients for
large Gross Domestic Product (GDP) reaching plants and their functions cannot be replaced by other
13.14% in 2017. In Indonesia, the area of oil palm elements so plants cannot grow normally if they are
plantations in 2017 according to the Central Statistics not available in sufficient quantities in the soil. The
Agency (BPS) in 2018 is estimated to reach more than essential nutrients studied in this research are calcium,
twelve million hectares, so to increase productivity magnesium, and sulfur. Sudradjat (2016) states that
requires proper fertilization. Fertilization on oil palm the main macro-nutrients needed by oil palm plants
plants aims to provide nutrient needs for plants so that are nitrogen, phosphorus, potassium, and magnesium.
plants can grow well and be able to produce optimally While the micro-nutrient that often become obstacles
and produce good quality oil (Adiwiganda and are copper, iron, manganese, and boron.
Siahaan 1994). To get the right fertilization results, Chlorophyll greatly affects the level of light
fertilizer recommendations for the palm oil reflectance in leaves. The reflectance of plants in
commodity are needed. Fertilization visible light (red, green, blue) (400-700 nm) and NIR
recommendations are to apply fertilization in a (700-900 nm) are strongly influenced by chlorophyll
manner and dose that have been determined in the and leaf cell structure (Stark et al. 2006). The light

127
Suyuthi, M., Seminar, K. and Sudradjat, .
Estimation of Calcium, Magnesium and Sulfur Content in Oil Palm using Multispectral Imagery based UAV.
DOI: 10.5220/0009978700002833
In Proceedings of the 2nd SEAFAST International Seminar (2nd SIS 2019) - Facing Future Challenges: Sustainable Food Safety, Quality and Nutrition, pages 127-134
ISBN: 978-989-758-466-4
Copyright c 2022 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved
2nd SIS 2019 - SEAFAST International Seminar

reflectance pattern can be used to assess plant health 2 RESEARCH METHODS

conditions related to the available of calcium,
magnesium and leaf sulfur nutrients. In this research, 2.1 Time and Location
parrot sequoia cameras are used to capture the light
reflectance of the leaves. The Phantom 4 Pro drone The study was conducted from February 2019 to
with a Parrot Sequoia camera is used to take April 2019 at the IPB-Cargill Oil Palm Education and
multispectral images that can capture 4 color Research Plantation, Singasari, Jonggol, Bogor, West
spectrum bands (green, red, red edge, and near- Java with coordinates 060 28,319' South Latitude,
infrared). Table 1 shows the specifications of the 1070 01.103' East Latitude and located 116 m above
Parrot Sequoia camera. sea level. The research was also conducted at the
Bioinformatics Engineering Laboratory of
Table 1 Specifications of Parrot Sequoia cameras. Mechanical and Biosystem Engineering, Faculty of
Spesifikasi kamera Parrot Sequoia Agricultural Technology, IPB University and Testing
Pixel 1.2 mp (1280x960) Laboratory of the Department of Agronomy and
Field of View 61.9o Horticulture, Faculty of Agriculture, IPB University.
Green band 530 – 570nm
Red band 640 – 680 nm 2.2 Materials and Tools
Red Edge band 730 – 740 nm
Near Infrared band 770 – 810 nm The tool used in this research is a computer with i7
RGB 400 – 700 nm processor speed of 2.4 GHz and has 8 GB of RAM.
Sumber: Parrot SA (2017)
Software used to process data are Microsoft Excel,
Microsoft Word, QGIS 2.18, Pix4DMapper, Agisoft
Aerial photography using Unmanned Aerial PhotoScan, and Minitab 18. Primary data used in this
Vehicles (UAV) is an alternative technology to get study were leaf nutrition obtained from leaf analysis
more detailed, real-time, fast and cheaper data testing at the Testing Laboratory of the Department of
(Shofiyati, 2011). UAV is a flying robot with remote Agronomy and Horticulture, Faculty Agriculture,
control that is able to carry payloads according to its Bogor Agricultural University. Another primary data
purpose and designation. This drone is capable of used is aerial photography from a 0.076 resolution
carrying cameras to photograph and record and can be Parrot Sequoia multispectral camera with the help of
flown to reach certain locations by remote control by a Phantom 4 Pro drone.
pilots. Many advantages if monitoring is carried out
with UAVs, including low investment and
2.3 Research Stages
operational prices, fast and flexible information
acquisition times, and information generated can be
The research began with the preparation phase of the
more detailed than satellite data. In addition, the UAV data consisting of leaf sampling, leaf sample
in transition flies under the cloud so that its image is coordinate points and Ground Control Point (GCP),
cloud free when compared to satellite imagery which
and drone image capture. Then the leaf samples were
depends more on atmospheric conditions (Dony analyzed at the Laboratory to determine the nutrient
2014). content used as the dependent variable. The drone
This research was conducted to estimate the
image data obtained must be carried out in the pre-
calcium, magnesium, and sulfur macro-nutrient data processing stage. This stage consists of photo
content of oil palm plants quickly and accurately by stitching, georeferencing, and digitization of garden
utilizing phantom drones 4. Taking images using
boundaries and canopy of sample plants. The results
multispectral parrot sequoia cameras that have high of the pre-processing data are in the form of reflectant
resolution. As well as leaf analysis to determine the
values in each sample plant and will be used as an
nutrient content of calcium, magnesium, and sulfur in
independent variable. After that the data analysis is
oil palm. The results of the image and nutrient content processed by multiple linear regression with stepwise
are correlated and interpreted in the form of a model.
methods using independent variables and dependent
variables. The results of the regression model are then
displayed in the form of an nutrient estimation layer.

128
 Estimation of Calcium, Magnesium and Sulfur Content in Oil Palm using Multispectral Imagery based UAV

2.3.1 Data Preparation Stage to be used as a reference point when uniting images
into a complete map. (Adillah 2018). A tool to draw
In the data preparation stage, several things are GCP coordinates can be by using a handheld GPS.
carried out, namely literature study, method
determination, discussion, and data collection. After 2.3.2 Pre-processing Stage
the research planning is done, the data collection
stage is carried out. The data collection phase is done The objectives of the pre-processing stage such as
by three things, namely sampling, drone imagery, and normalization and noise reduction are to produce
GCP. Leaf sampling is carried out in a spread where clean and ready-to-use data (Nanda et al. 2019; Nanda
the sample plants are proportionally determined. et al. 2018a; Nanda et al. 2018b). At this stage, the
Guidelines for sampling leaves on oil palm plants stitching and georeferencing process is performed
based on Winarna et al. (2005) where the leaves used using pix4Dmapper software. The stitching process is
as the main sample must meet a number of provisions, basically a combination or combination of two or
namely they are not the mains of inserts, grow more different images to create or form one image
normally, do not lie adjacent to roads or ditches/rivers, called a panorama (Kale and Singh 2015). Before
do not coexist with insertion trees and are not attacked stitching, georeferencing needs to be done, namely
by pests or diseases. the process of giving geographic references to raster
Leaf samples taken were leaves from the 17th or images that do not yet have a coordinate system
midrib. According to Chapman and Gray (1949) in reference. The coordinate reference used is WGS 84 /
Pahan (2006) said that the leaves of the 17th midrib UTM zone 48S with EPSG: 32748.
are the most sensitive leaves because they show the Then do bordering of the garden and canopy of the
greatest difference in nutrient levels. In addition, sample plants done using QGIS software. The results
nutrient status on the 17th leaf has a better correlation of the bordering of the sample plant canopy are used
to crop production when compared to other younger to extract the reflectance value in the form of a digital
leaves. The leaves of the 17th midrib are taken by six number at each pixel. The tool used to extract these
leaflets (three strands on the left and three strands on values is Zonal Statistics on the Raster menu.
the right at the meeting point of the two sides of the
midrib). Leaves that have been taken are stored in 2.3.3 Data Analysis Stage
envelopes that have been labeled according to the
location of the sample. The selected sample plants Estimation using multiple linear regression analysis
were given raffia to indicate the tree was a sample with the stepwise method is performed by Minitab
plant. Then the sample plants are marked on GPS which aims to determine the factors that effect and
which will be used to correct the geometry between make a estimation model of calcium, magnesium, and
the map and the image results. Samples that have been sulfur nutrition. At this stage two types of variables
obtained were analyzed for nutrients of calcium, are needed, namely the independent variable and the
magnesium, and sulfur in the Testing Laboratory of dependent variable. The independent variable
the Department of Agronomy and Horticulture. contains the reflectance value data which is the
Multispectral image capture using a Phantom 4 average value of the drone image in the digitizing
drone with a Parrot Sequoia camera. Drone flight attribute of the sample plant. While the dependent
planning automatically uses the Pix4D Mapper variable contains the actual calcium, magnesium, and
application which is adjusted to the taking land area sulfur nutrition data from the 17th midrib nutrient
and hours of time that can be taken by the drone. analysis results in the sample plants.
According to Kasih (2012) the optimum shooting is The model obtained is then evaluated to see the
done in the morning because the effect of reflected strength of the estimator model. The method used to
light from the sun is still weak. Besides the wind evaluate the model is the Mean Absolute Percentage
speed in the morning still tends to be low, thereby Error (MAPE). MAPE shows how big the difference
reducing the risk of UAV shake while shooting which between the actual results and the predicted results.
can cause poor quality captured images. Table 2 shows the criteria used to decide the
GCP determination aims to reduce errors or predictive power of an estimator model.
changes in position when integrating the results of the
image into a map that can make the data change. GCP
can be in the form of objects, buildings, or forms of
certain locations that can be clearly seen on the image
so that the coordinates of the object can be measured

129
2nd SIS 2019 - SEAFAST International Seminar

Table 2: Criteria for estimating the strength of the model. plantations and sample canopy are done manually
MAPE (%) Power of prediction using the add feature tool after creating a new
<10 Very Good polygon layer in QGIS software. Figure 1 and Figure
10 – 20 Good 2 show the results of bordering the sample plant
20-50 Moderate canopy digitized to extract the average value of the
>50 Bad digital number (DN) at each pixel. The extraction
Source: Wang et al. (2012) process is carried out on all layers of stitches using
Zonal Statistics in QGIS. Extraction results can be
After that, the nutrient estimation layer is made seen in appendix 1.
which contains the estimated nutrient value for each
pixel. The estimated nutritional value is the
application of the model obtained using the raster
calculator tool in the QGIS application. Then the
estimate value is classified based on the criteria in
Table 3.

Table 3: Nutrition concentration of 17th fronds on the age

of oil palm is more than 6 years.
Nutrition Unit Deficiency Optimum Excess
Ca %DM <0.25 0.50-0.75 >1.00
Mg %DM <0.20 0.25–0.40 >0.70
S %DM <0.20 0.25-0.35 >0.60 Figure 1: Results of the borderingon oil palm plantations.
Source: Von Uexkull and Fairhurst 1991

The results of the nutrient estimation layer that has

been classified will be tested for accuracy to see the
size of the miscalculation of nutrient estimates so that
it can be determined the percentage of accuracy of
nutrient estimation. Accuracy tests are performed
using a confusion matrix.

3 RESULT AND DISCUSSION

Overall data were collected on 6 February 2019 at the
Figure 2: Results of the bordering canopy of sample plants.
Cargill IPB Oil Palm Education and Research
Plantation. The oil palm plantation, which was
cultivated starting in 2012, has an area of 59 hectares Nutrition estimation is represented in the form of
which is divided into 5 plantation blocks. There are 3 mathematical models. Modeling uses the Minitab 18
types of data taken in this study, namely the 17th application with stepwise multiple linear regression
frond leaf sample, the coordinates of the sampling analysis. Modeling involves 23 samples from 30
location and GCP, and the multispectral image. samples. While the remaining 7 samples are used for
Retrieval of data is done in a day because the data model validation. The independent variable used is
needed is dynamic. Retrieval of image data using a the average DN value of each pixel obtained from the
DJI Phantom 4 drone with a Parrot Sequoia digitized of the plant canopy sample and the
multispectral camera. The drone was flown in dependent variable is the nutrient content of calcium,
accordance with the Parrot Sequoia handbook, which magnesium, and sulfur laboratory test results. The
is 110 meters above the ground, with a drone speed of results of the stepwise multiple linear regression
10 m/s and a pause time of 2.1 seconds for shooting. model are as follows:
The next step is the pre-processing of data in the form Ca= 0.994 + 0.00723 GREEN- 0.00863 RED EDGE
of stitching, georeferencing and bordering. Before Mg= 0.693 + 0.00531 RED- 0.00541 RED EDGE
that stage aerial photo printing results are selected S = -0.222 + 0.00338 RED EDGE
first using Agisoft PhotoScan then stitching and The model obtained is then predicted with stat>
georeferencing using QGIS. Restrictions on oil palm Regression> Regression> Predict. Table 4 shows a

130
 Estimation of Calcium, Magnesium and Sulfur Content in Oil Palm using Multispectral Imagery based UAV

comparison of actual and predicted nutrition. optimum conditions, light blue at above optimum
Evaluate the model using Mean Absolute Percentage conditions, and dark blue at excess nutrient conditions.
Error (MAPE). Evaluation of the model is done to see Figure 3 shows the results of the calcium nutrient
how strong the estimation of the model is made. The estimation layer, Figure 4 for magnesium nutrition,
results of the model evaluation obtained the value of and Figure 6 for sulfur nutrition.
MAPE in the Ca model, Mg model, and the S model The result of nutrient estimation layer is
respectively 22.72%, 11.62%, and 46.5%. Based on performed by accuracy test using a confusion matrix.
the strength estimation criteria of the model according In the confusion matrix the predicted nutrition of the
to Wang et al (2012) in table 2, the evaluation results model results is compared with the actual nutrition of
show that the Ca and S models are categorized as the lab test results. From the results of the accuracy-
medium and the Mg model is categorized as good. test using a confusion matrix, Overall Accuracy
The model obtained is then interpreted in the form of obtained in the nutrients of calcium, magnesium, and
an estimated nutrient layer that has been processed sulfur was 66.6%, 63.3%, and 36.7%. According to
and classified according to table 3. Classification of Jaya (2015) a good accuracy value is an accuracy
nutritional assumptions is presented in the form of a value that has reached a score of> 85%, so the
color that matches the nutrient content. The red color estimation of calcium and magnesium nutrients has
indicates that the oil palm tree on the grid lacks moderate accuracy and for sulfur has poor accuracy.
nutrients, yellow under optimal conditions, green at

Table 4: Comparison of actual nutrition and prediction of the 17th midrib leaf laboratory results.
Magnesium (%DM) Kalsium (%DM) Sulfur (%DM)
Sample
Actual Prediction Actual Prediction Actual Prediction
1A 0.36 0.296 0.51 0.643 0.15 0.231
1B 0.34 0.296 0.69 0.594 0.17 0.282
1C 0.19 0.304 0.53 0.616 0.17 0.266
1D 0.32 0.316 0.82 0.668 0.13 0.193
1E 0.32 0.284 0.69 0.622 0.43 0.264
2A 0.32 0.383 0.39 0.621 0.18 0.193
2B 0.42 0.356 0.59 0.642 0.14 0.218
2C 0.32 0.378 0.69 0.649 0.13 0.184
2D 0.41 0.390 0.91 0.658 0.40 0.167
2E 0.39 0.353 0.79 0.651 0.17 0.203
2F 0.35 0.351 0.68 0.638 0.17 0.233
2G 0.34 0.379 0.67 0.683 0.07 0.156
2H 0.44 0.386 0.42 0.668 0.09 0.181
3A 0.40 0.357 0.48 0.632 0.46 0.220
3B 0.38 0.370 0.68 0.603 0.40 0.265
3C 0.36 0.424 0.82 0.671 0.16 0.125
3D 0.30 0.375 0.79 0.615 0.14 0.236
3E 0.40 0.385 0.61 0.633 0.22 0.192
3F 0.41 0.389 0.66 0.661 0.34 0.157
3G 0.47 0.355 0.62 0.647 0.23 0.202
3H 0.45 0.416 0.55 0.629 0.20 0.182
3I 0.41 0.461 0.60 0.582 0.23 0.218
3J 0.49 0.458 0.51 0.583 0.28 0.221
3K 0.44 0.410 0.70 0.570 0.19 0.286
3L 0.33 0.375 0.67 0.591 0.36 0.263
4A 0.48 0.362 0.56 0.608 0.26 0.247
4B 0.50 0.423 0.72 0.634 0.10 0.147
4C 0.37 0.387 0.87 0.625 0.17 0.169
4D 0.43 0.413 0.45 0.619 0.15 0.165
4E 0.40 0.382 0.52 0.577 0.15 0.243

131
2nd SIS 2019 - SEAFAST International Seminar

Figure 3: The calcium nutrient estimation layer.

Figure 4: The magnesium nutrient estimation layer.

132
 Estimation of Calcium, Magnesium and Sulfur Content in Oil Palm using Multispectral Imagery based UAV

Figure 5: The sulfur nutrient estimation layer.

4 CONCLUSION show the accuracy of estimation of calcium and

magnesium nutrition including moderate grade, while
The nutritional estimation model was made by using the estimation of sulfur nutrition is poor.
stepwise multiple linear regression analysis. The
independent variable used is the average
multispectral reflectance value in the form of a digital REFERENCES
number in the canopy of the sample plant and the
dependent variable used is the nutrient content of Adillah, Y. (2018). Analisis Pertumbuhan dan
calcium, magnesium, and sulfur on the 17th midrib Produktivitas Padi dengan Nilai Indeks Vegetasi
leaf obtained from analysis in the testing laboratory Menggunakan Kamera Multispektral UAV (Analysis of
of the Department of Agronomy and Horticulture, Growth and Productivity of Rice with Vegetation Index
Values Using UAV Multispectral Cameras.) Master.
IPB. The model obtained from multiple regression
IPB University.
analysis for calcium nutrition is Ca= 0.994 + 0.00723 Badan Pusat Statistik. (2018). Statistik Kelapa Sawit
*GREEN - 0.00863*RED EDGE, for magnesium Indonesia 2017 (Indonesian Oil Palm Statistics 2017).
nutrition is Mg= 0.693 + 0.00531*RED - 0.00541 Indonesia: Badan Pusat Statistik.
*RED EDGE, and for sulfur nutrition is S= -0.222 + Adiwiganda, R. and Siahaan, M.M. (1994). Kursus
0.00338*RED EDGE . Evaluate the model using Manajemen Perkebunan Dasar Bidang Tanaman
Mean Absolute Percentage Error (MAPE). MAPE (Basic Plantation Management Course in Plant Fields).
values in the Ca, Mg, and S models are respectively Medan, Indonesia: Lembaga Pendidikan Perkebunan
22.72%, 11.62%, and 46.57%. The results of the Kampus Medan, p. 68.
Chapman, G.W. and Gray, H.M. (1949). Leaf Analysis and
model evaluation show that the strengths of the Ca
the Nutrition of Oil Palm. Annals of Botany, 13(4), pp.
and S models are in the moderate category and the Mg 415-433.
model is in a good category. While the accuracy of Dony, K. (2014). Teknologi Akuisisi Data Pesawat Tanpa
the nutrient estimation layer using confusion matrix Awak dan Pemanfaatannya untuk Mendukung Produksi
obtained Overall Accuracy value for estimating Informasi Penginderaan Jauh (The Unmanned Aerial
calcium nutrition by 66.7%, magnesium nutrition by Vehicles Data Acquisition Technology and Its
63.3%, and sulfur nutrition by 36.7%. The results

133
2nd SIS 2019 - SEAFAST International Seminar

Utilization to Support the Production of Remote Sudradjat. (2016). Kelapa Sawit: Peningkatan
Sensing Information). Inderaja, 5(7), pp. 24-31. Produktivitas(Palm Oil: Increased Productivity).
Goh, K.J., Hardter, R. (2003). General Oil Palm Nutrition. Bogor, Indonesia: IPB Press, p. 68.
In: Fairhust T.H., Hardter, ed., Oil Palm: Management Von Uexkull, H.R. and Fairhurst, T.H. (1991). Fertilizing
for Large and Sustainable Yield. Switzerland: PPI., pp. for High Yield and Quality. Basel, Switzerland:
191-230. International Potash Institute.
Gerendas, J. and Heng, A. (2010). Oil Palm Fertilization – Wang, F., Hsiao, Y. and Chang, K. (2012). Combining
Sharing Some Perspectives. Plantation Industry: Diffusion and Grey Models Based on Evolutionary
Competitive Strategies In Achieving A Sustainable Optimization Algorithms to Forecast Motherboard
Future. Selangor, Malaysia, pp. 1-6. Shipments. Journal of Mathematical Problems in
Jaya, I.N.S. (2015). Analisis Citra Digital: Perspektif Engineering, pp 1-10.
Penginderaan Jauh untuk Pengelolaan Sumberdaya Winarna, Sutarta, E.S. and Sugiono. (2005). Pedoman
Alam (Digital Image Analysis: A Remote Sensing Pengambilan Contoh Daun dan Tanah pada Tanaman
Perspective for Natural Resource Management). Bogor, Kelapa Sawit (Guidelines for Sampling Leaves and
Indonesia: IPB Press. Soils in Palm Oil Plants). Medan, Indonesia: Pusat
Kale, P. and Singh, K.R. (2015). A Technical Analysis of Penelitian Kelapa Sawit.
Image Stitching Algorithm. International Journal of
Computer Science and Information Technologies, 6(1),
pp. 284-288.
Kasih, L.S.B. (2012). Penentuan Iluminansi dan Ketinggian APPENDIX
Terbang Pesawat Optimum untuk Pemetaan Tingkat
Warna Daun Padi (Determination of Optimum Aircraft Digital number values of digitization of the sample plant
Illuminance and Altitude for Mapping the Color of Rice canopy.
Leaves). Bachelor. Bogor, Indonesia: IPB University.
Sample Green NIR Red Red Edge
Nanda, M. A., Seminar, K. B., Nandika, D., & Maddu, A.
(2019). Development of Termite Detection System 1A 97.696 150.290 55.234 126.499
Based on Acoustic and Temperature Signals. 1B 99.226 162.430 60.269 128.360
Measurement, 147, 106902. 1C 107.011 158.916 62.020 128.823
Nanda, M. A., Seminar, K. B., Nandika, D., & Maddu, A. 1D 100.055 149.172 54.509 120.953
(2018a). A Comparison Study of Kernel Functions in 1E 101.391 162.627 55.288 125.934
the Support Vector Machine and Its Application for 2A 92.608 131.776 71.301 127.583
Termite Detection. Information, 9(1), 5. 2B 110.319 123.704 75.585 139.960
Nanda, M. A., Seminar, K. B., Nandika, D., & Maddu, A. 2C 103.668 124.700 73.264 132.114
(2018b). Discriminant Analysis as a Tool for Detecting 2D 104.833 123.975 73.732 129.730
the Acoustic Signals of Termites Coptotermes 2E 107.398 130.823 70.143 133.121
curvignathus (Isoptera: Rhinotermitidae). International 2F 118.497 135.772 75.499 137.183
Journal of Technology, 9(5): pp. 840-851. 2G 112.261 130.371 69.891 125.834
Pahan, I. (2006). Panduan Lengkap Kelapa Sawit 2H 122.389 129.942 77.669 131.776
(Complete Palm Oil Guide). Jakarta, Indonesia: 3A 107.613 133.104 73.033 134.595
Penebar Swadaya. 3B 125.211 138.418 85.341 141.739
Parrot, S.A. (2017). Parrot Sequoia Capture the Iinvisible 3C 102.690 124.847 73.835 121.059
Monitor Your Crops. [Online]. Avaiable at: 3D 114.509 134.506 80.459 137.144
https://www.parrot.com/us/business-solutions/parrot-
3E 103.420 133.144 73.862 128.594
sequoia [Accessed 1 January 2018].
3F 101.884 131.221 69.520 123.551
Pusat Data dan Sistem Informasi Pertanian. (2014). Outlook
3G 105.717 133.299 69.451 131.241
Komoditi Kelapa Sawit (Palm Oil Commodity
Outlook). Jakarta, Indonesia: Pusat Data dan Sistem 3H 108.704 131.464 80.579 128.028
Informasi Pertanian Sekretariat Jenderal-Kementerian 3I 121.342 135.507 97.446 132.733
Pertanian. 3J 122.091 132.585 98.171 135.082
Shofiyati, R. (2011). Teknologi Pesawat Tanpa Awak untuk 3K 134.299 137.072 99.472 146.368
Pemetaan dan Pemantauan Tanaman dan Lahan 3L 114.651 134.315 84.555 141.575
Pertanian (Unmanned Aerial Vehicles Technology for 4A 111.116 137.428 77.785 136.967
Mapping and Monitoring Crops and Agricultural Land). 4B 90.891 131.026 72.818 119.278
Informatika Pertanian, 20(2):,pp. 58–64. 4C 79.805 129.379 66.129 122.606
Starks, P.J., Zhao, D., Phillips, W.A. and Coleman, S.W. 4D 85.741 126.997 73.204 123.770
(2006). Development of Canopy Reflectance 4E 96.247 141.433 77.157 131.028
Algorithms for Realtime Prediction of Bermuda Grass
Pasture Biomass and Nutritive Values. J. Crop Sc,.
46(2), pp. 927-934.

134